CN110110661A - A kind of rock image porosity type recognition methods based on unet segmentation - Google Patents
A kind of rock image porosity type recognition methods based on unet segmentation Download PDFInfo
- Publication number
- CN110110661A CN110110661A CN201910375770.7A CN201910375770A CN110110661A CN 110110661 A CN110110661 A CN 110110661A CN 201910375770 A CN201910375770 A CN 201910375770A CN 110110661 A CN110110661 A CN 110110661A
- Authority
- CN
- China
- Prior art keywords
- image
- model
- unet
- rock
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of rock image porosity type recognition methods based on unet segmentation comprising following steps: S1, builds initial depth learning network model;S2, it obtains rock original image and carries out image cutting and image enhancement, obtain pretreated image data;S3, it obtains rock original image and carries out the artificial mark of aperture position and shape, the label image data after being marked;S4, One-Hot coding is carried out to the label image data after mark, the label data after being encoded;S5, the label data after pretreated image data and coding is trained initial depth learning network model as training sample, the model after being trained;S6, images to be recognized is identified using the model after training.Noise Resistance Ability of the present invention is strong, has generalization ability, hole accuracy of identification can be improved, and realize the identification of hole classification.
Description
Technical field
The present invention relates to a kind of blowholes to identify field, and in particular to a kind of rock image hole based on unet segmentation
Kind identification method.
Background technique
Rock porosity is to measure its index for keeping fluid capacity, and different types of blowhole has different
Feature, such as shale have intergranular pore, intragranular hole, organic matter hole, gap.The difference of pore character may finally cause by permeability
There are greatest differences for the development effectiveness of size control, therefore porosity type largely determines oil recovering efficiency.In recent years,
With the development of digital image processing techniques, a kind of common method of identification porosity type is ground using drill cores sample
Casting body flake shoots Slice Image under scanning electron microscope and handles image, extracts the feature of blowhole image, thus
Classification and Identification is carried out to it.
Traditional images dividing method carries out image segmentation by color, shape and Texture eigenvalue to image first, later
Subsequent work is carried out again.Such methods have biggish limitation, and this kind of algorithm generally has some standards artificially formulated, and
There is no learning ability, relatively good result can only be obtained for specific scene.More importantly such dividing method can only obtain
To hole shape feature and be unable to get the classification of hole, hole can only be carried out by artificial or sorting algorithm further
Classification, and Rock Matrix locating for hole classification and hole has larger association, therefore carries out classification effect to hole sample merely
Fruit is poor.Method based on traditional images segmentation has:
(1) based on the dividing method of threshold value: the dividing method based on threshold value is that the gray feature based on image is calculated or set
Fixed one or more gray threshold, then by the way that the gray value of pixel each in image to be compared with threshold value, to reach
One classification results based on division, and then complete the segmentation of image.The shortcomings that the method, is if foreground area in image
This method when perhaps color distinction is smaller close with the pixel grey scale of background area will appear over-segmentation or less divided.
(2) based on the dividing method at edge: the dividing method based on edge determines the edge in region, so by detection
Different classification is marked off according to edge afterwards, the precondition of this method is that region to be sorted has more apparent edge feature,
But poor result can be obtained by distinguishing the discontinuous situation in unobvious or edge for edge.
Summary of the invention
For above-mentioned deficiency in the prior art, a kind of rock image hole class based on unet segmentation provided by the invention
Type recognition methods solves the problems, such as that existing blowhole identification is difficult.
In order to achieve the above object of the invention, the technical solution adopted by the present invention are as follows:
There is provided a kind of rock image porosity type recognition methods based on unet segmentation comprising following steps:
S1, semantic segmentation model unet is built initially as network model, using convolutional neural networks as basic framework
Deep learning network model;
S2, it obtains rock original image and carries out image cutting and image enhancement, obtain pretreated image data;
S3, it obtains rock original image and carries out the artificial mark of aperture position and shape, the label figure after being marked
As data;
S4, One-Hot coding is carried out to the label image data after mark, the label data after being encoded;
Label data after S5, the coding for obtaining the obtained pretreated image data of step S2 and step S4 as
The initial deep learning network model that training sample builds step S1 is trained, the model after being trained;
S6, images to be recognized is identified using the model after training.
Further, step S1 method particularly includes:
Using semantic segmentation model unet as network model, addition after each convolutional layer of unet model is normalized into layer,
And zero padding operation is carried out in each convolutional layer of down-sampling layer;By the hole classification number of initial deep learning network model
5 are set as, setting learning rate adjustment mode is that adam optimizes gradient decline, and the picture number that every wheel iteration batch processing is arranged is
24, setting the number of iterations is 10000, and setting iteration precision is 0.0001.
Further, image cutting in step S2 method particularly includes:
By image change at the image of 512 × 512 or 256 × 256 pixel sizes.
Further, the specific method of image enhancement includes following sub-step in step S2:
Connected domain is less than a in S2-1, removal imagebThe hole of pixel size obtains the image of removal part hole;Wherein
B is the number of plies of pond layer during initial deep learning network model down-sampling;A is the nuclear parameter of pond layer;
S2-2, Fuzzy Processing is carried out using image of the low-pass filtering to removal part hole, the figure after obtaining Fuzzy Processing
Picture;
S2-3, noise is added to the image after Fuzzy Processing, and the image after addition noise is carried out 90 °, 180 ° respectively
It is rotated with 270 °, forms new image data, obtain expanding the image set after quantity.
The invention has the benefit that the method for the present invention can make full use of the advantage of semantic segmentation hole identification problem,
Hole classification identification end to end is realized, and on the one hand applicable scene is single for traditional images dividing method, can not handle more
Scape challenge;Another aspect conventional segmentation methods hole accuracy of identification is low, and segmentation effect is poor.It is proposed by the present invention to be based on unet
The rock image porosity type recognition methods of segmentation is suitable for more scene challenges, Noise Resistance Ability using deep learning frame
By force, there is generalization ability, hole accuracy of identification can be improved, and realize the automatic identification of hole classification.
Detailed description of the invention
Fig. 1 is flow diagram of the invention.
Specific embodiment
A specific embodiment of the invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific embodiment, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the column of protection.
As shown in Figure 1, should based on the unet rock image porosity type recognition methods divided the following steps are included:
S1, semantic segmentation model unet is built initially as network model, using convolutional neural networks as basic framework
Deep learning network model;
S2, it obtains rock original image and carries out image cutting and image enhancement, obtain pretreated image data;
S3, it obtains rock original image and carries out the artificial mark of aperture position and shape, the label figure after being marked
As data;
S4, One-Hot coding is carried out to the label image data after mark, the label data after being encoded;
Label data after S5, the coding for obtaining the obtained pretreated image data of step S2 and step S4 as
The initial deep learning network model that training sample builds step S1 is trained, the model after being trained;
S6, images to be recognized is identified using the model after training.
Step S1's method particularly includes: using semantic segmentation model unet as network model, by each volume of unet model
Addition normalization layer after lamination, and zero padding (padding) operation is carried out in each convolutional layer of down-sampling layer;It will be initially deep
The hole classification number of degree learning network model is set as 5, and setting learning rate adjustment mode is adam optimization gradient decline, if
The picture number for setting every wheel iteration batch processing is 24, and setting the number of iterations is 10000, and setting iteration precision is 0.0001.
Batchnorm layers can make final result more rapid convergence, and padding operation can be convenient to calculate, so that network layer is more clear
It is illustrated, and then unet model is made to retain all features when characteristic pattern is sheared, reduces the characteristic loss of down-sampling, Jin Erbao
More information are stayed, the effect of original unet model is improved.
Image is cut in step S2 method particularly includes: by image change at 512 × 512 or 256 × 256 pixel sizes
Image.
The specific method of image enhancement includes following sub-step in step S2:
Connected domain is less than a in S2-1, removal imagebThe hole of pixel size obtains the image of removal part hole;Wherein
B is the number of plies of pond layer during initial deep learning network model down-sampling;A is the nuclear parameter of pond layer;
S2-2, Fuzzy Processing is carried out using image of the low-pass filtering to removal part hole, the figure after obtaining Fuzzy Processing
Picture;
S2-3, noise is added to the image after Fuzzy Processing, and the image after addition noise is carried out 90 °, 180 ° respectively
It is rotated with 270 °, forms new image data, obtain expanding the image set after quantity.
In one embodiment of the invention, rock raw image data collection is zoomed in and out, make its size 512 ×
512 pixels are so as to the every workload for taking turns data processing of reduction model in the case where retaining sufficiently high pixel.The figure for having mark
As being fabricated to mask images by low-pass filtering, i.e., it is Rock Matrix that foreground part, which is hole background parts, different hole classes
Different colors can not be corresponded to, background parts are unified for black.Selected 400 random points change noise spot as noise spot
Pixel value so that it becomes white, can be enhanced the robustness of model.Remove connected domain in the present embodiment less than 16 pixel sizes
Hole, i.e., connected domain is become consistent black with background less than the color of the hole of 16 pixel sizes in the image after mask
Color.One-Hot coded treatment is carried out after converting numpy.array format for the image for removing part connected domain.
When carrying out the identification of hole classification, test image path and setting are kept in modification evalute.py first
Model parameter path runs evalute.py, exports the blowhole mask images of semantic segmentation model prediction, passes through Pixel-level
Other logic or operation merge the mask images of prediction with original image, finally obtain the blowhole classification identification figure of prediction
Picture, and then hole classification is obtained, complete identification.
In conclusion the method for the present invention can make full use of the advantage of semantic segmentation hole identification problem, realizes end and arrive
The hole classification at end identifies, and on the one hand applicable scene is single for traditional images dividing method, can not handle more scene challenges;
Another aspect conventional segmentation methods hole accuracy of identification is low, and segmentation effect is poor.Rock proposed by the present invention based on unet segmentation
The recognition methods of image porosity type is suitable for more scene challenges using deep learning frame, and Noise Resistance Ability is strong, has general
Hole accuracy of identification can be improved in change ability, and realizes the automatic identification of hole classification.
Claims (4)
1. a kind of rock image porosity type recognition methods based on unet segmentation, which comprises the following steps:
S1, semantic segmentation model unet is built into initial depth as network model, using convolutional neural networks as basic framework
Learning network model;
S2, it obtains rock original image and carries out image cutting and image enhancement, obtain pretreated image data;
S3, it obtains rock original image and carries out the artificial mark of aperture position and shape, the label image number after being marked
According to;
S4, One-Hot coding is carried out to the label image data after mark, the label data after being encoded;
Label data after S5, the coding for obtaining the obtained pretreated image data of step S2 and step S4 is as training
The initial deep learning network model that sample builds step S1 is trained, the model after being trained;
S6, images to be recognized is identified using the model after training.
2. the rock image porosity type recognition methods according to claim 1 based on unet segmentation, which is characterized in that institute
State step S1's method particularly includes:
Using semantic segmentation model unet as network model, by addition normalization layer after each convolutional layer of unet model, and
Zero padding operation is carried out in each convolutional layer of down-sampling layer;The hole classification number of initial deep learning network model is arranged
It is 5, setting learning rate adjustment mode is that adam optimizes gradient decline, and the picture number that every wheel iteration batch processing is arranged is 24,
It is 10000 that the number of iterations, which is arranged, and setting iteration precision is 0.0001.
3. the rock image porosity type recognition methods according to claim 1 based on unet segmentation, which is characterized in that institute
State what image in step S2 was cut method particularly includes:
By image change at the image of 512 × 512 or 256 × 256 pixel sizes.
4. the rock image porosity type recognition methods according to claim 1 based on unet segmentation, which is characterized in that institute
The specific method for stating image enhancement in step S2 includes following sub-step:
Connected domain is less than a in S2-1, removal imagebThe hole of pixel size obtains the image of removal part hole;Wherein b is first
The number of plies of pond layer during beginning deep learning network model down-sampling;A is the nuclear parameter of pond layer;
S2-2, Fuzzy Processing is carried out using image of the low-pass filtering to removal part hole, the image after obtaining Fuzzy Processing;
S2-3, to after Fuzzy Processing image be added noise, and by be added noise after image carry out respectively 90 °, 180 ° and
270 ° of rotations, form new image data, obtain expanding the image set after quantity.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910375770.7A CN110110661A (en) | 2019-05-07 | 2019-05-07 | A kind of rock image porosity type recognition methods based on unet segmentation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910375770.7A CN110110661A (en) | 2019-05-07 | 2019-05-07 | A kind of rock image porosity type recognition methods based on unet segmentation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110110661A true CN110110661A (en) | 2019-08-09 |
Family
ID=67488616
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910375770.7A Pending CN110110661A (en) | 2019-05-07 | 2019-05-07 | A kind of rock image porosity type recognition methods based on unet segmentation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110110661A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110596166A (en) * | 2019-09-11 | 2019-12-20 | 西京学院 | Method for identifying type and content of oil-gas reservoir space |
CN110717060A (en) * | 2019-09-04 | 2020-01-21 | 平安科技(深圳)有限公司 | Image mask filtering method and device and storage medium |
CN110910444A (en) * | 2019-11-14 | 2020-03-24 | 中国科学院力学研究所 | RES scale shale equivalent three-dimensional pore parameter rapid extraction method |
CN111507966A (en) * | 2020-04-17 | 2020-08-07 | 无锡雪浪数制科技有限公司 | Composite material pore detection method based on UNET deep network |
CN111783815A (en) * | 2020-01-15 | 2020-10-16 | 中国地质调查局发展研究中心 | Multi-scale sampling and input method of rock stratum deep learning identification model |
CN112686259A (en) * | 2020-12-16 | 2021-04-20 | 中国石油大学(北京) | Rock image intelligent identification method and device based on deep learning and storage medium |
CN112700395A (en) * | 2019-10-17 | 2021-04-23 | 中国石油天然气股份有限公司 | Micro pore identification method based on U-Net full convolution depth neural network |
WO2021082918A1 (en) * | 2019-10-28 | 2021-05-06 | 上海悦易网络信息技术有限公司 | Screen appearance defect detection method and device |
CN112907498A (en) * | 2019-11-18 | 2021-06-04 | 中国商用飞机有限责任公司 | Pore identification method, device, equipment and storage medium |
CN113743227A (en) * | 2021-08-06 | 2021-12-03 | 山东大学 | Rock mass fracture intelligent identification method and system based on proximity photography |
CN116122802A (en) * | 2022-12-23 | 2023-05-16 | 中国科学院地质与地球物理研究所 | Drilling logging feature extraction method and system based on Unet double-channel output |
CN116309567A (en) * | 2023-05-17 | 2023-06-23 | 西南石油大学 | Shale electron microscope pore intelligent recognition method for small sample |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108346145A (en) * | 2018-01-31 | 2018-07-31 | 浙江大学 | The recognition methods of unconventional cell in a kind of pathological section |
CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
CN109190682A (en) * | 2018-08-13 | 2019-01-11 | 北京安德医智科技有限公司 | A kind of classification method and equipment of the brain exception based on 3D nuclear magnetic resonance image |
CN109284780A (en) * | 2018-09-10 | 2019-01-29 | 中山大学 | Ore mineral image automatic identification and classification method |
CN109712111A (en) * | 2018-11-22 | 2019-05-03 | 平安科技(深圳)有限公司 | A kind of cutaneum carcinoma category identification method, system, computer equipment and storage medium |
-
2019
- 2019-05-07 CN CN201910375770.7A patent/CN110110661A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108346145A (en) * | 2018-01-31 | 2018-07-31 | 浙江大学 | The recognition methods of unconventional cell in a kind of pathological section |
CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
CN109190682A (en) * | 2018-08-13 | 2019-01-11 | 北京安德医智科技有限公司 | A kind of classification method and equipment of the brain exception based on 3D nuclear magnetic resonance image |
CN109284780A (en) * | 2018-09-10 | 2019-01-29 | 中山大学 | Ore mineral image automatic identification and classification method |
CN109712111A (en) * | 2018-11-22 | 2019-05-03 | 平安科技(深圳)有限公司 | A kind of cutaneum carcinoma category identification method, system, computer equipment and storage medium |
Non-Patent Citations (1)
Title |
---|
郭文惠: "基于深度学习的岩石组分研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110717060A (en) * | 2019-09-04 | 2020-01-21 | 平安科技(深圳)有限公司 | Image mask filtering method and device and storage medium |
CN110717060B (en) * | 2019-09-04 | 2023-08-18 | 平安科技(深圳)有限公司 | Image mask filtering method, device and storage medium |
CN110596166A (en) * | 2019-09-11 | 2019-12-20 | 西京学院 | Method for identifying type and content of oil-gas reservoir space |
CN112700395A (en) * | 2019-10-17 | 2021-04-23 | 中国石油天然气股份有限公司 | Micro pore identification method based on U-Net full convolution depth neural network |
WO2021082918A1 (en) * | 2019-10-28 | 2021-05-06 | 上海悦易网络信息技术有限公司 | Screen appearance defect detection method and device |
CN110910444B (en) * | 2019-11-14 | 2022-12-09 | 中国科学院力学研究所 | RES scale shale equivalent three-dimensional pore parameter rapid extraction method |
CN110910444A (en) * | 2019-11-14 | 2020-03-24 | 中国科学院力学研究所 | RES scale shale equivalent three-dimensional pore parameter rapid extraction method |
CN112907498B (en) * | 2019-11-18 | 2024-04-30 | 中国商用飞机有限责任公司 | Pore identification method, device, equipment and storage medium |
CN112907498A (en) * | 2019-11-18 | 2021-06-04 | 中国商用飞机有限责任公司 | Pore identification method, device, equipment and storage medium |
CN111783815A (en) * | 2020-01-15 | 2020-10-16 | 中国地质调查局发展研究中心 | Multi-scale sampling and input method of rock stratum deep learning identification model |
CN111783815B (en) * | 2020-01-15 | 2024-02-13 | 中国地质调查局自然资源综合调查指挥中心 | Multi-scale sampling and input method for rock stratum deep learning recognition model |
CN111507966B (en) * | 2020-04-17 | 2024-02-06 | 无锡雪浪数制科技有限公司 | Composite material pore detection method based on UNET depth network |
CN111507966A (en) * | 2020-04-17 | 2020-08-07 | 无锡雪浪数制科技有限公司 | Composite material pore detection method based on UNET deep network |
CN112686259A (en) * | 2020-12-16 | 2021-04-20 | 中国石油大学(北京) | Rock image intelligent identification method and device based on deep learning and storage medium |
CN112686259B (en) * | 2020-12-16 | 2023-09-26 | 中国石油大学(北京) | Rock image intelligent recognition method and device based on deep learning and storage medium |
CN113743227A (en) * | 2021-08-06 | 2021-12-03 | 山东大学 | Rock mass fracture intelligent identification method and system based on proximity photography |
CN116122802A (en) * | 2022-12-23 | 2023-05-16 | 中国科学院地质与地球物理研究所 | Drilling logging feature extraction method and system based on Unet double-channel output |
CN116122802B (en) * | 2022-12-23 | 2023-07-14 | 中国科学院地质与地球物理研究所 | Drilling logging feature extraction method and system based on Unet double-channel output |
CN116309567A (en) * | 2023-05-17 | 2023-06-23 | 西南石油大学 | Shale electron microscope pore intelligent recognition method for small sample |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110110661A (en) | A kind of rock image porosity type recognition methods based on unet segmentation | |
CN110070552A (en) | A kind of rock image porosity type recognition methods based on semantic segmentation | |
CN111047551B (en) | Remote sensing image change detection method and system based on U-net improved algorithm | |
CN107480707A (en) | A kind of deep neural network method based on information lossless pond | |
CN107644426A (en) | Image, semantic dividing method based on pyramid pond encoding and decoding structure | |
CN105069779B (en) | A kind of architectural pottery surface detail pattern quality detection method | |
CN101398894B (en) | Automobile license plate automatic recognition method and implementing device thereof | |
CN109558806B (en) | Method for detecting high-resolution remote sensing image change | |
CN104050471A (en) | Natural scene character detection method and system | |
CN107506822A (en) | A kind of deep neural network method based on Space integration pond | |
CN112750106B (en) | Nuclear staining cell counting method based on incomplete marker deep learning, computer equipment and storage medium | |
CN107480649A (en) | A kind of fingerprint pore extracting method based on full convolutional neural networks | |
CN114331869B (en) | Dam face crack semantic segmentation method | |
CN107066916A (en) | Scene Semantics dividing method based on deconvolution neutral net | |
CN112819096B (en) | Construction method of fossil image classification model based on composite convolutional neural network | |
CN113255837A (en) | Improved CenterNet network-based target detection method in industrial environment | |
CN112837344A (en) | Target tracking method for generating twin network based on conditional confrontation | |
CN110717921B (en) | Full convolution neural network semantic segmentation method of improved coding and decoding structure | |
CN110472634A (en) | Change detecting method based on multiple dimensioned depth characteristic difference converged network | |
CN107958219A (en) | Image scene classification method based on multi-model and Analysis On Multi-scale Features | |
CN107480723A (en) | Texture Recognition based on partial binary threshold learning network | |
CN108052932A (en) | One kind blocks adaptive face identification method | |
CN103325123A (en) | Image edge detection method based on self-adaptive neural fuzzy inference systems | |
CN116993737B (en) | Lightweight fracture segmentation method based on convolutional neural network | |
CN114926386A (en) | Method and system for automatically segmenting leucocyte nucleoplasm based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190809 |
|
RJ01 | Rejection of invention patent application after publication |